Update on the AI Act: the EU reached an agreement on the ‘final’ text
After the first proposal in April 2021, the approval of the draft Act in June 2023 and the trilogue negotiations (European Commission, European Parliament and Council) last month, an agreement has been reached by the EU.
As mentioned in earlier insights on our website, the AI Act will apply to providers and deployers users of AI systems established in the EU as well as in third countries if the output of the AI system is used in the EU.
The EU agreed to prohibit the following AI systems posing unacceptable risks and potential threats to citizen’s rights:
- Biometric categorisation systems that use sensitive characteristics (e.g. political, religious, philosophical beliefs, sexual orientation, race) (with exception of such use for law enforcement purposes);
- Untargeted scraping of facial images from the internet or CCTV footage to create facial recognition databases;
- Emotion recognition in the workplace and educational institutions;
- Social scoring based on social behaviour or personal characteristics;
- AI systems that manipulate human behaviour to circumvent their free will;
- AI used to exploit the vulnerabilities of people (due to their age, disability, social or economic situation).
Clear obligations were established for high-risk AI systems such as biometric identification used in public spaces for law enforcement purposes, AI systems used in critical infrastructures (e.g. healthcare), AI in education and vocational training, AI in safety components of products and AI in evaluation systems, law enforcement and migration and asylum. Comprehensive mandatory obligations were established such as fundamental rights impact assessments, conformity assessments, risk mitigation, data governance, documentation, transparency, robustness, accuracy and cybersecurity obligations.
Furthermore, the EU agreed on transparency obligations and other requirements for general-purpose AI (GPAI) and foundation models.
Non-compliance with the AI Act can lead to sanctions such as fines between 1,5% or 7% of global turnover, depending on the type of infringement and size of your company.
There are still some details to be finalised but it seems that the AI Act will come into effect in 2026.
Need to know more? Contact Nuans.
